I am wondering if there is a way to predict more than one test case with tk-instruct. Someting like in this format: Model definition Positive examples Negative examples Now predict the followings Input1 Input2 Input3